jump to navigation

Impact and Data to Growth (Notes from April 8 – 14, 2019) May 1, 2019

Posted by Anthony in cannabis, Digital, experience, finance, global, NLP, questions, social, Strategy, training, Uncategorized, WomenInWork.
Tags: , , , , , , , , , , , , , , , , , , , ,

I started watching Street Food on Netflix and in the Osaka episode, the chef makes a great claim that can work for today’s notes: “If you want to create your own current, you cannot live your life by going with the flow.” Granted, you can use the current as a guide, but to truly create something unique, you have to hop out of the path and try your own luck. Today, I was listening to an episode even with Keith Rabois on Invest Like the Best, and he’s a proponent of not making 10% decisions, but rather investing into 10x ones – the riskiest that can pay off are the ones that will be truly incremental. The 10% experiments may improve a bit, but won’t exponentially get you scaling.

I had a great mix of NLP / Machine Learning podcasts, social/responsible corporations like United by Blue and Everytable, to go with sales thought processes, data ethics, and finance starts on a global scale. Each person / founder / company tackling unique challenges based on their individual experience that got them to that point. How can you approach the problem? And more importantly, what’s the right way that your expertise leads you to a solution for this problem?

For some, it was how to release the stigma in the cannabis industry to expose people to the health benefits as we’re legalizing in more states? United by Blue’s founder wanted a truly sustainable business model that supported his beliefs in giving back. An expert data scientist by the name of Debbie continues to improve women relations in tech and data-related fields by 1) supporting others graciously and 2) providing, particularly Latin American women, the opportunity to see how her passion for learning sparked her adventurous career.

Hope you enjoy! Leave a comment or follow along!

  • Su Wang, Elisa Ferracane in Authorship Attribution, UT (Data Skeptic, 1/25/19)
    Link to ACLWeb for paper

    • Discourse units in addition to others
      • Rhetoric structure theory (RST) – 2 elementary clauses (as Elementary Discourse Units – EDU)
      • Relation is related by an ‘elaboration’ where the 2nd sentence elaborates on the previous sentence
      • Rows are sentence pairs and the cells show the relations between the 2 (1st, 2nd; 2nd, 3rd, etc…)
    • Plagiarism detection, authorship attribution as semantic inference (both authors as computational linguistic PhD)
    • Can be unsupervised (classification of text to an author style) or supervised (accuracy or how closely it matches an author – assign key)
    • For the paper, they looked at 9 texts via Project Gutenberg and did a CNN – high-level baseline
      • Had 2 months to get it to the next level, optimization – said that LSTM performed the best but too slow for translations or 1000s of words
      • CNN can be as good as LSTM or better depending on architecture
      • Tried grammatical matrix, columns are entity, rows as sentences – subject, object, other
    • Used dataset of 19 books and 9 authors as extension of prior state-of-the-art paper
      • IMDB as another dataset – short texts with many authors (tried to do with Twitter but can’t get structure/sentence)
      • Initial data set was ~15% more accurate (99.8%)
      • 98.5% accurate for extended novel classification ~50 texts – SVM did well also of about 84-85% (more data may allow them to be more acc)
    • Looking at the different types of features – RST was more sophisticated in that the models did better in all experiments
      • Could embed or use the one factor as a distribution over other set of features
    • For IMDB dataset, discourse features nearly didn’t help – too short to establish structure
    • Human as the ‘gold standard’ but certainly not perfect. Authorship probably different task, though.
      • Would require expertise on the authors’ part. Machine can pick up on far more patterns.
    • Next for him – semantic narratives and story salads (grant via DARPA?)
      • Coherent narratives, shuffle the sentences and reconstruct the story.
  • Sylvia Wehrle, founder CEO of June CBD Apothecary (Wharton XM)

    • Talking about the difference between CBD, THC and other strands
    • Humans as growing up with various forms of hemp oil – additive and purposeful for our evolution
    • Using the appropriate properties to go through benefits – getting the common questions out of the way
  • Donald Robertson (@donjrobertson), author of How to Think Like a Roman Emperor: Stoic Philosophy of Marcus Aurelius (Wharton XM)
    • Book discussing difference between stoic and Stoic, cynic and Cynic, etc…
    • Calm and indifference is different than how it may have been perceived
  • Sam Polk (@sampolk), author and CEO of EveryTable (Wharton XM)

    • Sustainability at Feast (prior company)
    • Using Feast as test-tasters for EveryTable menu / offerings
    • EveryTable as sustainable, healthy food for people in an affordable way
      • Restaurants with partnerships of cities/areas that match the pricing (Santa Monica different than Watt or Compton)
      • Can order on app and go pick up meal for < $8 – able to do this with scale – try to ensure this early
    • Rolling out BlueApron-style weekly meals at the same price as in store
    • Corporate offerings where they have EveryTable coolers / fridges that take a credit card / payment and can pull out your order
  • Brian Linton, United by Blue founder CEO (Wharton XM)

    • Originally moved from Singapore / Asia, went to college in Michigan – boredom satisfaction with sales
      • Started with ‘guady’ jewelry that was travel-related (tourist-style jades, emeralds, etc…) that he would source from home
      • Travel down to Florida / other areas and sell to region
    • Believed in doing good, so he would donate ~5% of all proceeds to ocean conservation – realized this wasn’t sustainable
      • Random donations of $1000s or %’s
    • Finally started United by Blue to develop the sustainable business model and what he believed in
  • Right Way to Get Your First 1000 Customers with Thales Teixeira (@thalesHBS), associate professor at HBS (HBR IdeaCast #676, 4/2/2019)
    • Startups failing because they try to emulate successful disruptive biz and scale instead of learning about initial customers
    • First customers are more than the money, word-of-mouth, R&D and free feedback
    • Etsy, Amazon, Netflix, Uber had no new technology (just finally had the map to see if there were cars coming)
      • Etsy went to craft fairs to recruit sellers, who then attracted more buyers
      • Pinterest tried to create a culture initially to set the tone for quality
      • AirBnb was awful initially in NY, so the founders wanted to find out – places were great but pictures were awful
        • Rented a nice camera and offered to take the pictures to improve the ones on the listings
    • What is the primary driver of value to the customers to deliver? How does technology play a role in this?
      • AirBnb had 1 engineer (founder) for a long time – increase the utilization of an expensive asset
        • Hid the options initially – didn’t have much inventory so they would email / find out and then get back to customers
        • Show availability – needed to stay in a house in the places
    • Technology is the enzyme / enabler of the start-up or experience and acquire the customers to purchase the product
      • People that like smaller companies, try new things, explore products and tell them
    • Unlocking the Customer Value Chain (Thales’ book)
  • Critical Thinking in D/S – Debbie Berebichez (@debbieberebichez) (DataFramed #58, 3/25/19)
    • Debbie is a physicist, TB host and CDS at Metis in NY (first Mexican woman to get a PhD in Physics from Stanford)
      • Promoting women in STEM, especially hispanic women
    • Metis is a data science teaching company as an arm of Kaplan in NY, San Francisco, Seattle
    • Did 2 postdocs around Columbia before going to Wall Street to work as a quant – but money wasn’t the only motivator, so she left
    • At Cambridge, she remembered speaking about Astronomy 101 as her first intro to physics class – was on 2 years of scholarship
      • She took a walk with her friend Rupesh and said that she was crying – “I just don’t want to die without trying physics.”
      • Passion drew attention and professors – offered her for a 2 year physics degree (skip first 2 if she could pass a test with complicated derivatives)
        • Had 2+ months to learn calculus, basics to mechanics and more – passed her test (9am to 9pm)
    • Mentioned going into high school to discuss data science – class was doing coding/SQL/data look on animals
      • Had 1 group that was looking over turtles – couldn’t answer the units for weight (triple digits) – not lbs, but grams
      • How this made sense – how to piece together reasoning / bias – how needed this skill was
      • Not bothering to check outliers or some data was exhibiting – why do we do it all?
      • Danish astronomer built and designed 1000 stars, which wasn’t much, but Newton and Kepler, Copernicus all derived theories from
    • Large datasets vs small datasets – insight more important vs size (big data as sometimes unnecessary)
    • Feynman quote about fooling ourselves – bias that we create.
    • History of Statistics – Stiegler, normal distribution and derivation of central limit theorem by Gauss and Laplace (1809 with Jupiter’s motion around sun)
    • With her bootcamp – she wants to attack the question of using the right algorithm and how to analyze the problems at hand
      • How to choose a data project in what you’re interested in – madewithmetis on Metis site
    •  Singular value decomposition (SVD) and reducing dimensionality, worked with Genentech founder – healthy DNA vs patient’s DNA and cancer
      • Reducing dimensions to the ones that were most relevant – NLP also
    • Think deeply, be bold, help others – Grace Hopper celebration talk
  • Dean Oliver (@deano_lytics), Data Analytics (Wharton Moneyball)

    • Talking about how far behind NFL is behind NBA in tracking
    • There are people doing video for football, but not much – not widespread
      • Position groups will gain entirely different/new insights into how they’re playing
  • Cordasco Financial Network Planning + Sri Thiruvadanthal (Behind the Markets, Jeremy Schwarz)
    • Discussion of hedging dollar vs not – if hedging, probably wise to diversify with global
      • If not hedging, then europe may not be as great
    • Current markets say that liquidity isn’t as high with central banks, stocks start to couple and lose diversification / value
      • Decoupling early on in cycles
    • Relative value may be fine but not absolute for the dollar compared to other currencies
  • Jeppe Zink, GP at Northzone (20min VC 087)

    • Invested in Spotify, Bloglovin, TrustPilot with focus on SaaS, fintech, mobile
    • Worked at Deutsche Bank as analyst in corporate finance, tech banker – left with 90% of team
      • Convince bank by buying principal investments before IPO in late 1990s – worked out
    • European cycles of tech – 100mln to 3bn people online, digital increase and telecom infrastructure
      • First VC firms in existence were doing integrated buyout model, which failed initially – too transaction focus
      • VCs have the talent that’s aligned with the founders now – 90% of VC firms that existed in 2000 had died in 2002
    • 10 year cycles where the great companies withstand, others don’t
    • Stage agnostic for them, series A to D rounds
      • Nordic companies of unicorns for what he has had success with
      • Europe as dropping trade barriers initially and in the 90s, broadband and smart phone starts (Nokia, Ericsson)
    • Has offices in the north for Northzone but he makes it up every other week or so
    • Try to emulate the start-up and have hunger/ambition always
      • Not trying to stagnate – venture capital vs patient (he thinks impatient is better – learn through failure and testing)
      • How fast can you learn to level up and deliver the best product? Continuous measurements, KPIs.
      • For Jeppe – momentum in product development
    • Most intrigued by fintech investing – Peter Thiel as one of his favorites
      • Most recent company was CrossLend – consumer lending with European bank lending
      • Book: Startup Growth Engines as collection of random founders and interviews


No comments yet — be the first.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: